Fails to parse lambda in PEP 695 TypeVar bound

[15r10nk]

Describe the bug

The following code can not be parsed/formatted by black:

class name_5[name_4: lambda: something]:
     pass

❯ black -t py312 bug.py
error: cannot format bug.py: Cannot parse for target version Python 3.12: 1:21: class name_5[name_4: lambda: something]:

Oh no! 💥 💔 💥
1 file failed to reformat.

but it can be parsed by cpython.

python3.12 scripts/min_code.py

Environment

• Black's version: current main (00c0d6d)
• OS and Python version: Linux/Python 3.12

Additional context

The bug was found by pysource-codegen (see #3908)

[MeGaGiGaGon]

I did some digging, and this is a parser issue, it tokenizes the same as with the builtin tokenizer module. The error happens while parsing the lambda token, but I couldn’t go further than that with my usual technique since it’s just an error with the generated abstract machine having no valid transition states. I looked at the grammar.txt file, but at least the part about typeparams looks to match what’s in the python docs, so no clue.

[JelleZijlstra]

This turned out to be a simple fix, I had used the wrong expression kind in the grammar.

It also incorrectly rejects these in TypeVar defaults (new in 3.13), and in addition to lambdas it rejects if ... else, and, or, not, and comparisons (== etc.). Fortunately none of these make semantic sense here.